| As an efficient and clean energy,the installed scale of wind power is con stantly expanding at home and abroad.However,due to the random and inter mittent characteristics of wind energy,large-scale integration into wind power brings great challenges to the peak regulation and frequency regulation of the power grid,and the contradiction between the rapid development of wind pow er and the safe and stable operation of the power system is becoming increasin gly prominent.Therefore,high-precision wind farm power prediction can prov ide the basis and guarantee for the safe and stable integration of wind power in to the grid.At present,the research on wind power prediction is mainly predic ted by physical methods and statistical methods,the prediction accuracy of the existing single model is poor,the prediction accuracy of the combined model has been improved,but the effect of improvement is not significant,and the es tablishment of wind power error correction model is relatively small,and the d istribution law is unstable.In order to solve the above problems,this paper tak es Longyuan Group’s wind farm as a case,applies the monitoring data of wind turbine monitoring data and acquisition(SCADA)system in the power plant t o predict the wind power in the power plant,and focuses on the comparative a nalysis and testing of the accuracy of the variational mode decomposition com bination model to predict wind power power.The research of the paper has cer tain practical significance.Firstly,the Raida criterion and interpolation method are used to eliminate and fill the abnormal variables,and then the random forest(RF),categorical r egression tree method(CART)and maximum information coefficient(MIC)a re used to rank the importance of the composite variables and construct a new data set.On this basis,the autoregressive differential moving average model(ARIMA)is used to predict the wind power of wind power,and the simulation results show that the ARIMA model can fully extract the available informatio n of the remaining terms.At the same time,the short-term prediction of wind power is based on three different prediction models of support vector machine(SVR),random forest(RF)and long short-term memory neural network(LST M),and it is found that LSTM has the highest prediction accuracy.Improved particle swarm(IPSO)is used to optimize the parameters of LSTM,verify its effectiveness with actual data,and further improve the prediction accuracy.Fi nally,multiple subsequences and residual sequences are used to reduce the vol atility of the data,IPSO-LSTM prediction is established for the separated mod al components,and then the ARIMA model is used to correct the residual sequ ence,and the prediction results of each subsequence are superimposed to obtai n the final prediction results.The simulation results show that the ARIMA mo del can correct the error generated by the prediction model itself,and verify th at the VMD-IPSO-LSTM-ARIMA model has better prediction accuracy and c an adapt to the situation with higher prediction accuracy. |
